ToolsCompaniesAbout

Deep Agents

by LangChain@langchainSan Francisco, United StatesNo ratings yet

Open-source agent harness for long-running, multi-step tasks

Visit langchain.com
Tool type
Open-source agent harness
Made for
Developers building autonomous agents
Runs on
Your code, or LangSmith-managed
Self-host the SDK or use Managed Deep Agents
Last checked
29 Oct 2026
4 product sources

From @LangChain, about Deep Agents

Recent important posts · features

Open on X

What it does

Deep Agents is an open-source agent harness from LangChain. You give it a model, a set of tools and a system prompt, and it runs an agent that can plan a task, read and write files in a virtual filesystem, call tools (including MCP servers), run code in a sandbox or interpreter, and hand work to subagents. Output is whatever the agent produces, such as research results, code changes or other task results, along with traces if you connect LangSmith.

Developers use it through the Deep Agents SDK, creating an agent with a single function call and choosing a model from providers such as Google, OpenAI, Anthropic, OpenRouter, Fireworks, Baseten or Ollama. It differs from a bare agent framework by bundling context management (history summarisation, offloading large tool results, prompt caching), skills, long-term memory, human approval steps and filesystem permission rules. Task planning is optional. A related terminal coding agent, dcode, is built on the SDK.

It suits developers building autonomous agents for research, coding or other long workflows who want defaults instead of assembling everything on LangGraph. You can self-host the open-source SDK or use Managed Deep Agents on LangSmith. You need a model provider account and code to configure the agent. It is not a no-code product, and managed hosting, tracing and sandboxes are billed through LangSmith.

What it can do

  • Planning tools for breaking tasks into steps and tracking progress
  • Virtual filesystem with pluggable backends and path-level permission rules
  • Subagents running in isolated context windows, synchronous or async
  • Context management: summarisation, offloading large results, prompt caching
  • Skills and long-term memory loaded on demand
  • Human-in-the-loop approval and interrupts

Works with

  • LangSmith
  • Model Context Protocol (MCP)
  • LangChain
  • LangGraph

Example workflows

Research agent with subagents
Inputs: a research question, a chosen model and search tools. Steps: the agent plans subtasks, spawns subagents for independent topics, and saves intermediate notes to the virtual filesystem. Output: a compiled report file plus a trace in LangSmith.
Coding task with approval gates
Inputs: a repository task and a sandbox backend. Steps: the agent edits files and runs tests via the execute tool, pausing for human approval at critical points. Output: modified code and test results.

Before you choose

  • Task planning is opt-in from v0.7, so enable it if your workflow needs it
  • Filesystem permissions do not apply to sandbox backends, which can run arbitrary commands
  • Model and LangSmith hosting costs are separate from the free open-source SDK

Listed under

Get started

Needs a model provider and some code

  1. Install the Deep Agents SDK and review the Quickstart in the docs
  2. Pick a model provider and supply your API credentials
  3. Call create_deep_agent with your tools and system prompt, then invoke it; optionally enable LangSmith tracing
Start from GitHub or docs

Plans and prices

The open-source harness can run in your own environment. Model-provider and infrastructure charges remain separate; optional LangSmith services have their own pricing.

Deep Agents SDK (open source)—
  • Self-hosted in your own process or container
  • Bring your own model provider

Source: docs.langchain.com/oss/python/deepagents/overview, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

Ratings and reviews

No ratings for Deep Agents yet.

Be the first to rate it

What’s changed

Free account

See what’s changed at Deep Agents

Launches, price moves and pivots, each with a date and a source. Free with an account.

Signals

Free account

See Deep Agents’s signals

Launch traction, hiring and release activity, tracked over time. Free with an account.

Data and security

Free account

See how Deep Agents handles your data

Training on your files, compliance, hosting and the models behind it. Free with an account.

Compare with similar tools

Free account

Compare Deep Agents side by side

Prices, platforms and features against similar tools, with sources. Free with an account.

Where this comes from

How we research
Company website
langchain.com: product, pricing and policy pages
YC directory mirror
Batch and company name
Social accounts
Accounts the company links, and its posts on X
Human review
Checked by a person
knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.